Abstract: Semantic annotation of Web data is a key step for Web information extraction. The goal of semantic annotation is to assign meaningful semantic labels to data elements of the extracted Web object. It is a hot research topic that has gained increasing attention all over the world in recent years. Conditional random fields are the state-of-the-art approaches taking the sequence characteristics to do better labeling. However, traditional conditional random fields can not simultaneously use existing Web databases and logical relationships among Web data elements, which lead to low precision of Web data semantic annotation. To solve the problems, this paper presents a constrained conditional random fields (CCRF) model to annotate Web data. The model incorporates confidence constraints and logical constraints to efficiently utilize existing Web databases and logical relationships among Web data elements. In order to solve the problem that the Viterbi inference approach of traditional CRF model can not simultaneously utilize two kinds of constraints, the model incorporates a novel inference procedure based on integer linear programming and extends CRF to naturally and efficiently support two kinds of constraints. Experimental results on a large number of real-world data collected from diverse domains show that the proposed approach significantly improves the accuracy of semantic annotation of Web data, and lays a solid foundation for Web information extraction.